Robilar

I have received the EVGA 780i A1 Motherboard. My understanding is that there will also be a T1 release but will as with the 680i series be identical with the exception of warranty bundled accessories.

Given the success of my Asus P5N32-E Motherboard Overclocking Guide I have decided to add a new thread.

http://enthusiast.hardocp.com/articl...50aHVzaWFzdA==

HardOCP Review

http://www.guru3d.com/article/nforce...li_review/482/

Review of the XFX version


http://www.techreport.com/articles.x/13790

Another reference review

A few things I've learned so far:

9.64 Chipset drivers available (dated March 18)

http://www.nvidia.com/object/nforce_winxp_9.64.html

(thanks to Tae)

Use P08 bios.

http://www.evga.com/forums/tm.asp?m=552263

P08 Updates:

* Improved nTune compatibility
* Fixed restoring disabled device states & fan speeds after suspend/resume
* CPU Microcode updates for Penryn CPUs
* Fixed PCIe compatibility issues
* Improved Fan Controls
* Improved Memory Compatibility


Ntune Beta version 6.0 is designed for the 780i and is very slick. Thanks to Nuclearjock for pointing out that this software will make your computer shut down very slowly but otherwise works fine.


The stock NB fan is way too loud. Syrillian made a great suggestion below (with a pic) about using an Antec Spot Cool (they cost about $12). I ordered one myself to give it a go. Anything will be better than this stock fan.

USB motherboard headers, Firewire Header and Front Panel connectors are in a terrible location if you use SLI (Especially if you have a removable motherboard).

I have also noticed in some initial testing that my current cpu needs less vcore to reach OC that I benched on my prior 680i boards. For 3.724 for example, I needed about 1.44 vcore to be stable (before vdroop). With this board I was able to do an orthos blend (Gave it an hour or so) at 1.41 vcore and it was stable. Not a huge difference but considering its the exact same chip it helps.


Courtesy of Ericeod, he was able to confirm that a single video card must be placed in the top slot and provided a screenshot.

http://www.neowin.net/news/software/...tools-600-beta

Ntune Beta 6.0 Link (for XP)


http://www.lavalys.com/beta/everestu...wel3fkdm5c.zip

Here is a link for the Everest Ultimate Beta that works as well




Vdroop Pencil Mod:

http://www.xtremesystems.org/forums/...3&postcount=16

Link to Joemaniaci's 780i heatpipe removal guide:

http://www.overclock.net/nvidia/3038...-here-now.html


Here is an FTP link for beta chipset drivers from EVGA (apparently resolves SATA DVD issues)

XP 32 bit only!

ftp://ftp.evga.com/driver/9.46_SATA.zip

Attached Thumbnails

         

Robilar

Here are my bios settings using a Q6600 GO at 3.2 Ghz and 3.6 Ghz

CPU Config:

Disable C1E State

Disable Execute Disable Bit

Disable Virtualization Technology

Standard CMOS:

Disable 1.44 Floppy (Unless you actually have one and need it for RAID drivers)

Advanced Bios:

I disable all boot devices except hard drive and disable boot other devices. This speeds boot up time and also in case you leave a memory stick by accident in the system prevents your system from taking time to try to boot from it.

Full Screen Logo: Enable

Advanced:

Disable HPET

Voltages:

Set Memory to 2.1 V (whcih is the default for my OCZ ram)

CPU Voltage:

For 3.2 I set it to: 1.3625

For 3.6 I set it to: 1.48125

(These will vary based on your chips' VID and stepping)

Other voltages:

(All set manually do not leave them on auto)

CPU FSB: 1.4v

Memory: To whatever your ram needs

Nforce SPP: 1.4v

Nforce MCP: 1.525v

HT Nforce SPP: 1.25V


FSB/Memory:

Set Memory clock mode to Linked & Synchronous


FSB/QDR: Set to 1600 (400 FSB)

For cpu multiplier, for 3.2 to 8x, for 3.6 to 9x.

Leave Memory speed at 800mhz

Memory Timings:

4-4-4-8

1T


System Monitor:

Dynamic Fan Control set to Manual

PNP/PCI:

Set initial display to PCIEx (default is PCI)

Integrated Peripherals:

Turn off Onchip IDE Channel 0

MAC Config Turn off MAC01

IEEE Controller to disable

HD Audio to Disable

Onboard FDC Controller to Disable

Serial port to Disable


Please note these settings reflect my own setup and preferences.

Attached Thumbnails

         

PizzaMan

GTL testing and tweaking for all 7 series nVidia boards. 750i, 780i and 790i

This guide is for those who have already read the Intel Overclocking Guide and/or understand the basics of overclocking and stress testing.

What is GTLREF?
Source
The above source is the basis for this guide. It's the guide I used to learn this tweaking method. This guide is to simplify the method and make it easier for overclockers to understand and implement.

How will GTL adjustments help my overclock?
Proper testing and adjustments of the GTL lanes will give you more overclocking potential with lower VTT/FSB voltage. It will give your system the stability it needs to achieve higher clocks without having to increase VTT/FSB voltage to an uncomfortable level.

Here are the GTL assignments:
FOR QUAD CORES:
Lane 0 is for Core 0,1 Data Bus (die 0)
Lane 1 is for Core 2,3 Data Bus (die 1)
Lane 2 is for Core 0,1 Address Bus (die 0)
Lane 3 is for Core 2,3 Address Bus (die 1)

FOR DUAL CORES:
Lane 0 is for Core 0,1 Data Bus (die 0)
Lane 1 is for Core 0,1 Address Bus (die 0)

As you can see each lane is not for each core. The lane assignments really are not that important. What's important is the testing. For quads you will need to test all four lanes. For duals, just lanes 0 and 1 need to be tested. Though, this thread is mostly for quads. Most duals can overclock easily without having to adjust GTLREF.

780i and 750i boards will be able to adjust between +0mv and +160mv. Where as, 790i boards can adjust between -155mv and +155mv.

First, you need to use the knowledge you already have about overclocking and start overclocking your CPU and board. You need to raise the FSB and Ghz as high as you can and maintain stability. If you start testing the GTL lanes from stock you will likely find that you will have to go back and do them over again. Which you may have to do anyway, but doing it at stock is not needed. It's preferred to leave all the lanes set to +0mv if you can. If not, try using auto. Make sure your overclock is 100% stable before starting your testing. I personally recommend using more then one stress test. 12+ hours Prime95, 2+ hours OCCT small data set, and 100 loops of a Linpack test(IBT, Linx or OCCT's Linpack). No single stress test trumps all others. Each test will test your CPU and board differently. It's very important that you start with a rock solid 100% stability. The slightest instability could skew your results and make all the time you spent testing the lanes worthless.


So now that you have it stable and all lanes are set to hopefully "+0mv". If not, that's OK. Just leave the lanes where they are until it's time to test that lane. For 65nm quads you will likely be around the 3.2-3.4Ghz range when you start having to increase a lot of VTT/FSB voltage. 45nm quads will likely be around 3.5-3.8Ghz.

Start with lane 0. Increase lane 0 by +10mv. Test it each time you bump it. I personally ran 20 loops of IBT. Keep bumping it +10 until your OC is unstable. Let's say you where able to take lane 0 up from +0mv to +120mv and it still be stable, any higher and it's unstable. Now you take the middle of 0 and 120, which is 60mv. This is where you are going to leave lane 0.

Let's say, for example, you started with lane 0 at +90mv as a result of OCing with the lanes set to auto and the board preset it to +90mv. You can start by either raising or lowering the value. Say you can increase it to +150mv and lower it to +30mv. Any lower or higher and its unstable. Now you will set that lane for +90mv, as that is the middle point between the two values. Repeat process for each lane. It will take a while, maybe even days, but after which you should be able to lower your FSB voltage and increase the CPU's stability. If you have a 790i board you can continue to lower below 0 to -155mv. From what I've seen thus far, most the 790i boards adjustments will end with negative adjustments rather then positive like the 750i and 780i boards.

You most test each lane independently. You cannot make adjustments on more then one lane at a time and maintain an accurate test.

After running your tests, before you start increasing FSB and overclocking more try lowering your VTT/FSB voltage and stress testing again. You should find that you can run it stable with lower vFSB. Vcore will not change, as GTL is a signal and doesn't directly effect vcore.


Here's an example of my Q6600 when I adjusted it at 3.2Ghz with 1600QDR.

Lane 0 +80mv (everything between +0 and +160 passed)
lane 1 +60mv (everything between +0 and +120 passed)
lane 2 +55mv (everything between +0 and +110 passed)
lane 3 +80mv (everything between +0 and +160 passed)

After I did this testing I was able to lower my FSB voltage from 1.4v to 1.3v and then OCing to 3.6Ghz was nothing but adding vcore and NB voltage. Before testing, it took me 1.5v FSB and 3.6Ghz/1800QDR still wasn't stable.

Later after reaching 3.6Ghz, still with 1.3v vFSB at 1800QDR, I went back and tested them again.

Lane 0 +80mv (everything between +0 and +160 passed)
Lane 1 +50mv (everything between +0 and 100 passed)
Lane 2 +60mv (everything between +10 and +110 passed)
Lane 3 +80mv (everything between +0 and +160 passed)

If your able to continue to overclock your CPU you will find that you may need to go back and test the lanes as you go higher with your FSB. The margins will become smaller and smaller as the FSB increases. I would recommend testing again every 100 to 200 QDR.

1.3v VTT/FSB should be enough to take you to 1900-2000QDR. Those you see with 1.4v and higher are having to compensate for improper GTL adjustments.

Q&A
Q: Does increase the GTLREF effect temperature?
A: No, the GTL values are not actually increasing core voltage. Your temperatures will likely stay the same.

Q: Can I test all the lanes at the same time?
A: No, to do so will not give you the accuracy that you will want to gain the stability you are looking for.

Q: Will knowing which core is failing in Prime 95 help me determine which lane to adjust?
A: Yes, if you are in the process of moving GTL values and testing. No, you are just overclocking and hoping a random GTL adjustment will make your system stable. Reason, if you are just overclock and not in the process of testing GTL then a failed core in Prime95 might not be an error in that particular core. You could have an error in the L2 cache that is shared with all cores. I wouldn't rely on Prime95. It's better to rely on yourself by only adjusting one lane at at time.

If you have any questions or see any errors in this guide please feel free to PM me. I'm glad to help when I can. If anything is not clearly written I would like to know so that I can rewrite it so that it's easy to understand. That's my goal. I've tried to make this as simple as possible. You shouldn't have to know everything about GTLs to be able to test and tweak them.

Now, go clock it!

~PizzaMan



@Rodilar, feel free to add this to the 750i and 790i threads.

Robilar

Courtesy of the EVGA forums

780i boards and ESA Compliance:

ESA is a system monitoring standard that many of the top manufacturers have agreed to support because it benefits them all. Currently, products like Everest, RivaTuner, NVidia System Monitor and others all have to write their own interface code for the various sensor chips, components, etc. Another problem with the current state of things is that a company that makes performance components like power supplies might want to add an interface that would allow software to monitor temperature, current loads, voltages, etc. except that there would be no guarantee that anyone would support their interface.

With ESA, each party to the standard agrees that all power supplies will use the same monitoring interface, Graphics cards will use another standard interfaces, all fan controllers will have thier own standard interface and so on. This way as soon as a power supply manufacturer releases a new ESA product, it will immediately work with any ESA standard monitor.

There have been many similar standards in the past such as Microsoft's "Perfmon" interface, SNMP (Simple Network Management Protocol), Java's ARM (Application Response Measurement) and others. These types of standards are extremely good for folks like us that spend lots of money on hardware and just want it all to work together without it being a huge problem. Personally, I think it's going to be very cool to know exactly how much current the graphics cards are sucking up at various performance/temperature levels, how the 12v rails are holding up as maximum current draw is approached or many other such performance measurement combinations that should be possible once fully ESA-equipped computers are available.

The spring on the bottom of the EVGA 780i motherboard is simply attached to the primary SPP heatsink assembly on the top and is designed to ground it to the case. Think of this just like the rear I/O panel with the little tabs that touch the metal cases around the audio, network and USB connectors to ground them to the case. PC's are now running at frequencies that are not only much higher than AM, FM or TV, they are approaching RADAR frequencies so the RFI (Radio Frequency Interference) they generate can play havoc with lots of stuff. Its theorized that some of the newer ESA monitors in development right now have found that their frequency readings are jumping all over if that little spring isn't in place so its advised you to use it since there certainly isn't any down-side to having it in place.


Note: This spring is no longer attached on some of the newer board revisions. No explanation from EVGA but with or without seems to have no effect on the board.

Attached Thumbnails

         

nuclearjock

Rob,

A while back I posted this relating a bad experience with a friend's XFX 780i. Well it turned out to be a flaky board. His replacement board worked beatifully so I now own one as well.

Since both the EVGA and XFX boards are identical, are you encouraging feedback from XFX users in this thread as well??

__________________

RAFFY

---------------

HappyVirus

Well for newbies like me, posts like this would help me very much...You get an early thank you from me xD

RAFFY

----------------

Litlratt

Making definitive statements, regarding something you obviously have a limited knowledge of, is not recommended.

__________________

nuclearjock

People with QX9650's will be interested in results from this thread.

680i does not support Yorkfield quads, except for one DFI board.

__________________